1. Field of Invention
This invention relates to a method and system for reducing crosstalk in an optical communication network.
2. Description of Related Art
Fiber-optic communication networks are experiencing rapidly increasing deployment. Especially rapid is the growth of segments that carry multi-gigabit digital data on multiple wavelengths over a single fiber strand using wavelength division multiplexing (WDM). Increases in wavelength channel density (i.e., reduced channel spacing) and increases in the data rate carried on individual wavelengths lead to an increase in nonlinear crosstalk between channels. For passive optical fibers, the crosstalk mechanisms are cross-phase modulation, four-wave mixing, and Raman crosstalk. Further, active components, such as fiber-based or semiconductor based optical amplifiers will add cross-gain modulation.
The nonlinear crosstalk effects are additive to the overall interference level. The addition occurs in terms of each additional wavelength channel contributing a crosstalk component to the overall interference level. The additive effect also occurs in systems that have multiple optical links with intermediate optical amplification, such that each link also contributes an additional crosstalk component to the overall noise level. It is well known that the details of the bit pattern on each channel are important for the accurate estimation of the noise levels.
Due to the additive property of the crosstalk, whenever there is signal correlation among the wavelength channels, the crosstalk level will be maximized. In general, the data transported on each wavelength is expected to be random and not correlated. However, there are situations where data correlation exists. One situation involves transmission formats in which specific framing structures that define a specific protocol are used for data transport (e.g., SONET, SDH, Ethernet, ESCON, FiberChannel). A second situation involves default null data that is transmitted when external data is not present on a specific channel.
For example, FIG. 1 depicts two channels λ1 and λ2 carrying OC-192 SONET-framed data. SONET delineates its frames via two framing bytes, A1 (11110110) and A2 (00101000). The length of the framing is directly related to the signal rate. In the case of OC-192 data, the framing contains a 192-byte sequence of A1 bytes, followed by a 192-byte sequence of A2 bytes. Following the initial framing bytes is the payload (information to be transmitted) and additional framing bytes. Similarly, for OC-48 data, the framing contains a 48-byte sequence of A1 bytes followed by a 48-byte sequence of A2 bytes. As shown in FIG. 1, if the data on wavelengths λ1 and λ2 are launched simultaneously, the data patterns on each channel are highly correlated due to the presence of the SONET framing.
It is apparent that the A1 bytes present a pattern overweighted with ‘1’s, and A2 bytes are overweighted with ‘0’s. Due to channel synchronization, Raman crosstalk in the fiber depletes power from the short-wavelength channels during A1 bytes, and transfers this power to the A1 bytes of long wavelength channels. FIG. 2A illustrates the envelope of the power transferred from short wavelengths to long wavelengths during the A1 bytes of the SONET frame. Conversely, Raman-mediated power transfer is substantially eliminated during the A2 bytes where there is an average power increase in short-wavelength channels and a corresponding power dip in the long-wavelength channels (FIG. 2B). This transfer of power is due to high correlation and represents unwanted crosstalk between channels.